Zh Tw Humanizer — AI skill for Claude Code
繁體中文(台灣)去 AI 味 agent skill:59 種中文 AI 寫作特徵、兩岸用語在地化、簡轉繁修正、隱形字元清除,附機械檢查腳本與 eval 測試集.
How to install Zh Tw Humanizer
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open acchuang/zh-tw-humanizer and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Zh Tw Humanizer does
繁體中文(台灣)去 AI 味 agent skill:59 種中文 AI 寫作特徵、兩岸用語在地化、簡轉繁修正、隱形字元清除,附機械檢查腳本與 eval 測試集。Traditional Chinese (Taiwan) humanizer skill for Claude Code / Codex / Cursor / OpenCode / Gemini CLI.
Alternatives in AI
- Botmux — Bridge Feishu/Lark to AI coding CLIs — Claude Code, Codex, Gemini, OpenCode… every DM, group or topic spawns i 1.2k ★
- Claw Empire — Command Your AI Agent Empire from the CEO Desk — A local-first AI agent office simulator that orchestrates CLI 1.1k ★
- AI In The Terminal — Complete companion guide for NetworkChuck's 'AI in the Terminal' video - Gemini CLI, Claude Code, Codex, and o 855 ★
README
zh-tw-humanizer
[](#installation) [](SKILL.md) [](LICENSE)
**繁體中文版 README → [README.zh-TW.md](README.zh-TW.md)**
A portable agent skill that removes signs of AI-generated writing from Traditional Chinese text and localizes it to **Taiwanese Mandarin** — Taiwan vocabulary (影片/軟體/網路/資料, never 視頻/軟件/網絡/數據), Taiwanese tone, particles, and natural sentence rhythm. It never outputs simplified Chinese or mainland-style phrasing.
Plain Markdown, so it runs in any harness that supports skill-style instructions (Claude Code, OpenCode, pi, Cursor, etc.).
Based on [Wikipedia's "Signs of AI writing"](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing) guide and the zh-TW [AI生成文的特徵](https://zh.wikipedia.org/zh-tw/Wikipedia:AI%E7%94%9F%E6%88%90%E6%96%87%E7%9A%84%E7%89%B9%E5%BE%B5) essay, extended with Taiwanese-Mandarin localization rules (兩岸用語差異、語助詞、台式語氣).
What it covers
- 60 AI writing patterns in Chinese (4 categories): 假真誠開場白、心理諮商語氣、對仗句「不是 X 而是 Y」、概念名詞化「○○感/○○性」、空話動詞「提升/打造/賦能」、連接詞堆疊「此外/綜上所述/值得一提的是」、破折號濫用、金句堆疊、排比句、同義詞輪換、虛指權威、幻覺引用、假精確(無來源的精確數字)、過程敘事(「經過分析」「深入研究後發現」)、立場真空、公式化開場、解說導引句、假推論、說教深度腔、金句公式、戲劇性短句轟炸、勸誡反問收尾、粗體轟炸、emoji 堆疊、編號切碎段落、表格誤用、預告式導言、模板佔位文字、工具痕跡、聊天機器人痕跡…
- Chinese-specific tells English humanizers miss (new in 1.3.0): 翻譯腔句式(「最……之一」「當……的時候」「對 X 進行 Y」)、系動詞迴避(「作為/扮演著……的角色」撐胖一句「是」)、過度強調關注度(「引發熱議」「多家媒體報導」)、引用層破綻(死連結、DOI 檢查碼錯、access-date 早於發表日)、列表式行文與行內粗體標題、標題結構與 Markdown 破綻(跳級標題、多個 H1、
---分隔線、中文裡的彎引號)。 - New in 1.4.0: 模糊的關聯詞(「息息相關」「密不可分」)、空轉分析(分析的架勢,同義反覆的內容)、交付語的 AI 味(commit/PR 說明的「已保留原有資訊」「進行了若干優化」)、文風斷層偵測(同一篇裡人寫的段落與 AI 補的段落分開處理)、各家模型的殘留標記(ChatGPT/Gemini/Grok/DeepSeek/Perplexity)。
- New in 1.5.0: 虛假範圍「從 X 到 Y」(兩個無關詞硬湊成涵蓋感)、「最」字堆疊偵測、放寬對同義詞輪換的誤判(舊模型痕跡,新模型已弱,非母語作者也常見)、擴充「檢測指引」的誤判防線(完美文法、正式與口語混雜、單獨轉折詞都不算數)。
- **Prompt-injection
Related Skills
Pii Guard
繁體中文(台灣)個人資料去識別化工具,讓業務文件可以安全地送進 AI 處理 / Reversible Traditional-Chinese (Taiwan) PII de-identification for LLM
Humanizer Hk
去 AI 味 · Strip the AI tells out of written Chinese. Hong Kong Traditional. Claude Code skill + paste-anywhere
Bu Ketao
不客套 — Strip Chinese AI verbosity. 繁體中文 LLM 輸出壓縮規則集。~72% token compression, zero semantic loss.
Plain Language Skill
Rewrites convoluted AI output into plain language — separate rule sets for English and Traditional Chinese 把繞口
Memory Forge Rs
Stop resetting satisfying AI chats — edit the memory instead. Local session manager for Claude Code, Codex & O
Oc Claw
A desktop pet that monitors your AI coding agents (OpenClaw, Claude Code, Codex, Cursor, Gemini CLI, opencode,
Related Agents
Agent Of Empires
Claude Code, OpenCode, Mistral Vibe, Codex CLI, Gemini CLI Coding Agent Terminal Session manager via tmux and
Developer Overview
code-container (container) creates isolated Docker environments for AI coding harnesses (Claude Code, OpenCode
Consult Agent
Execute cross-tool AI consultations via Task spawning. Use when agents or workflows need a second opinion from